A NONPARAMETRIC VERTICAL MODEL:
AN APPLICATION TO DISCRETE TIME COMPETING
RISKS DATA WITH MISSING FAILURE CAUSES
Bonginkosi D. Ndlovu1
School of Mathematics, Statistics and Computer Science, University of KwaZulu-Natal,
Pietermaritzburg, South Africa
e-mail: bongi@dut.ac.za
Sileshi F. Melesse
School of Mathematics, Statistics and Computer Science, University of KwaZulu-Natal,
Pietermaritzburg, South Africa
Temesgen Zewotir
School of Mathematics, Statistics and Computer Science, University of KwaZulu-Natal, Durban,
South Africa
Discrete time competing risks data continue to arise in social sciences, education etc., where
time to failure is usually measured in discrete units. This data may also come with unknown
failure causes for some subjects. This occurs against a background of very limited discrete
time analysis methods that were developed to handle such data. A number of continuous
time missing failure causes models have been proposed over the years. We select one of
these continuous time models, the vertical model (Nicolaie et al., 2015), and present it as a
nonparametric model that can be applied to discrete time competing risks data with missing
failure causes. The proposed model is applied to real data and compared to the MI. It was
found that the proposed model compared favorably with the MI method.
Key words: Discrete Time Competing Risks, Missing Failure causes, Nonparametric Vertical
Model, Relative hazards, Total hazards.
1. Introduction
Discrete time competing risks data arise in survival analysis experiments when subjects are exposed
to multiple risks of failure and the time to failure evolves discretely. Time to failure
˜
T is measured in
discrete units if
˜
T ∈{1,..., q} for some positive integer q. This occurs when
˜
T is inherently discrete
or when
˜
T is originally continuous but observed failure times have been grouped into intervals. The
distinguishing feature of this data is an excessive number of event/censoring ties. In education,
for example, students may exit the system either by graduating or withdrawing. There are often
a substantial number of graduation ties because students can only graduate at the end of the year
or semester. Even though students can withdraw continuously throughout the year or a semester,
and in most instances informally, the authorities are able to ascertain a withdrawal if a student
1
Corresponding author.
MSC2010 subject classifications. 62-07, 62G05.
South African Statistical Journal
Vol. 54, No. 2, 231–241
httpsȷ//doi.org/10.37920/sasj.2020.54.2.7
© 2020 South African Statistical Association
231